Keith Burgun

Thoughts on Game Design

The Myth of Difficulty


“Game Difficulty”, as we’ve always thought of it, manifests in terms of questions like, “how hard should this game be” and features such as difficulty modes.  Unfortunately, I’ve found that this is a fundamentally flawed concept.  It doesn’t actually have any utility, except in systems that already are broken to begin with.

To clarify:  the myth is not that different games have different levels of difficulty; they absolutely do.  The myth is that we can freely scale that with difficulty modes or other solutions.


A Balanced Game

Instead of worrying about “difficulty”, I worry about “balance”.  People tend to think of balance most often in multiplayer games, but the same rules apply in single-player games.  I’ve written at length about the topic of balance for Gamasutra, but for now, it suffices to say that a game must be a tightly-woven fabric that is stable, even when vigorously tested.  When players play it, and do their best to abuse it, its fibers pull at each other in such a way that the attempt at an exploit is quickly countered;  not by some external dynamic-difficulty-adjustment meta-resource thing, but by the system itself.  Good gameplay design is self-correcting, but only when there is balance between all in-game actions and resources.

So, balance is very important.  If you don’t have balance, your game is dead.

How does this tie into the concept of “difficulty”?  Well, let’s look at what “game difficulty” normally refers to.


Game Difficulty

Usually, “difficulty” determines how “hard” it is to achieve the game’s goals.  Different people have different levels of skill at certain things, or different propensities for gaining skills, so it would certainly make sense if we could offer several different difficulty “modes” to different players.

We’ve always thought that indeed, this hard-ness can be scaled up and down simply by changing some variables:  how many monsters on a level, how many hit points things have, or tweaking other values.

The problem is that this is not possible in a balanced game.

If you have a balanced game, then you do not have the capability of tweaking any variables at all.  Doing so would necessarily imbalance your game.  Increasing the hit points of zombies in Castlevania, forcing you to hit them twice each instead of once with your whip dramatically changes the nature of how the game is played, has huge effects on secondary weapons, and way more.  The system known as Castlevania has an optimal configuration, a point of balance where the game is most interesting.  Moving outside that is necessarily making the game worse, no matter who is playing it.

An extreme example that can highlight this is Bethesda’s RPGs, wherein hit-point scaling is the accepted way of increasing game difficulty as the game goes on.  The problem is, having to hit a bear 50 times may indeed be “harder” in a way, but it’s vastly more boring.  This is an extreme example of how this kind of “scaling” does not work, because you’re moving away from that optimal configuration.

Designers do have control over the hardness of their game while they are designing/balancing it, but once they’ve finally reached a point of “this is now a balanced, interesting, engaging game” (NOTE: Getting to this point AT ALL is extremely difficult and most designers fail to do it even once), they don’t really have control over how hard or easy their game is.  This is a significant thing to realize!

In fact, the only way to create a new “easier” or “harder” game out of an existing balanced game is to take it apart, change some values, and then put it back together again.  Essentially, though, you’re going to have to balance this other gameplay mode.  It is probably most healthy to think of difficulty modes, then, as variants.  This makes sense, because if you changed some values and then rebalanced the game around that, the game is going to have a distinctly different character than it did before you started the re-balancing.

Half-Caveat:  Most videogames are not balanced to begin with, due to having far, far too much inherent complexity.  So, in these systems, you can indeed get away with having difficulty modes, but it’s quite like how you can get away with murder on the Titanic.


Caveat:  Difficulty in Puzzles

In Puzzles, difficulty makes perfect sense, because puzzles do not have to be “balanced”.  Puzzles can be linear, and the purpose of them is solution, so finding some kind of “imbalanced element” has limited effect on the system.

It’s also worth mentioning that many if not most single player videogames are puzzles of some sort;  very rarely are they games as I define the term.  However, most “obviously puzzle” videogames don’t have difficulty modes anyway.  It’s usually videogames that have some game-like elements to them — with “combat” and such — that have difficulty controls, and this is what I’m objecting to.


I hope that I’ve made my point clear:  you cannot scale difficulty in a balanced game.  Learning to play a game is learning a discipline, and not all disciplines are equally easy or hard to learn.  Some are harder than others, by their nature, and that’s OK.  Not all games need to be for all people.





Posted in: Balance, Game Design

  • Disquisitor Sam

    I get what you’re saying in this article. I think you kinda shot down your own point, though. I’d agree that most single player games are more puzzle-like than game-like, and difficulty settings are found almost exclusively in single player games. Therefore if difficulty makes perfect sense in a puzzle, then all these games aren’t wrong to add difficulty. The only testing you need to do for those is whether the solution is “possible” rather than whether or not the game works. Any design problems would lie elsewhere.

    One exception to what you described would be something like Starcraft or Civilization where the difficulty setting affects exclusively the strength of the AI. Sometimes the designer will cheat and allow the AI to break the rules at higher difficulties, and it highlights why a lot of games played vs. AI are spectacularly bad at teaching legitimate strategies for multiplayer. Now that I think of it, AI without some degree of randomness is still puzzle-like, so maybe it’s not so much of an exception after all.

    What kind of consideration are you giving towards difficulty and AI for Empire? Or do your foes follow the same rules and have the same situations that the player does?

  • Curly

    “Learning to play a game is learning a discipline, and not all disciplines are equally easy or hard to learn.”

    The trouble is, the typical modern singleplayer game is not designed to be replayed. The campaign of something like Gears of War is optimised for the person who sits down and plays through it once and then never touches it again. I’ll stick with FPSes as an example, since you referenced Doom. The difficulty levels in those games usually increase the damage dealt by enemies, reducing the player’s margin for error, or increase the number of enemies that spawn, forcing him to kill efficiently or be overwhelmed. They vary the challenge to the player along a single dimension – the test of his FPS control skills. The descriptions of the difficulty levels in Gears and games like it reflect this. They say things like “I can pull off a headshot” or “I’m new to shooters.”

    They don’t say “I’ve played this game before.” It’s a completely different balancing problem to challenge the player who knows the rules, knows the maps, and knows what he’s supposed to be doing. Shooters that try to cater to experienced players usually botch it horribly – they just take the same twitch skill difficulty knobs and turn them up to 11, changing tough opponents into one hit killers that have to be defeated with AI-exploiting tactics and save scumming. Done properly, a higher difficulty level would be a test of all the other skills the player is exercising while he’s playing, demanding more of his spatial awareness, tactical ability against increasingly complex configurations of enemies and terrain, flexibility in the face of changing circumstances, and so on. More bullet damage won’t get at any of that. The designers would need to take apart the game and rebuild it from the ground up.

    So I agree with your thesis. But again, most singleplayer games are single-use amusement park rides. For the game that assumes the player doesn’t know how to play, a difficulty setting that only asks “How fast are your reflexes?” or “How accurately can you aim?” and tweaks a few numbers in response doesn’t seem to be inherently balance-breaking. It usually feels like the same game.

  • keithburgun

    >>Therefore if difficulty makes perfect sense in a puzzle, then all these games aren’t wrong to add difficulty.

    Like I said, though – puzzle-ish videogames tend to not even have difficulty settings.

    It’s more of a question of games like Civilization, or actioney games like Ninja Gaiden or something, which actually DO have difficulty settings.

    >>What kind of consideration are you giving towards difficulty and AI for Empire?

    Foes do indeed follow the same rules/situations that the player does, but that still leaves questions for AI. Personally I’m actually kind of resentful of AI in videogames in general, because basically all it is is a crappy puzzle laid on top of a game. What I can tell you now is that the AI will be extremely simple; dumb, even, because surviving the system is what’s hard. Later, we’ll probably re-tweak the game for a multiplayer variant, which actually is probably the natural state for this game.

  • keithburgun

    >>The trouble is, the typical modern singleplayer game is not designed to be replayed.

    Yeah, well this either means that they are a puzzle (about solution) and therefore not affected by my point (as I said), OR they are a crappy game with little depth (which is a much larger problem).

    So yeah, a single-use amusement park ride obviously is not affected here. I would generally classify most of those as puzzles, though (if even execution puzzles).

  • Dasick

    About the worst thing a game can do is make the “AI” symmetrical to player. Asymmetry is the way to go. And the best way to make challenging AI is to give it no-brainer, direct abilities that still create gameplay for the player. Most monsters in Doom actually have this to some degree (except for the hit-scan zombies). And of course monsters in Auro are a great example of this :D

    What you can do is have AI have certain cards that are unavailable to the player, but which are unambiguous in their use, and make the player jump some hoops. Maybe like, have a lot of reaction cards (if player does X, do Y), and some really brute force things? Maybe the enemy has some sort of a reinforcement advantage, where they don’t lose when they lose 3 units, only when their line is damaged (but the player loses combat if she loses 3 monsters)

    Or you could even turn off other empires (or modify them to be unlike the player), and have the player deal with mostly the desolation monsters? And in the multiplayer you can tone down effects of desolation, requiring active action from player to start spawning monsters, so that you don’t have the ‘survive until your enemy dies off’ type of gameplay.

  • keithburgun

    Yeah, the AI will certainly have to be cheat-y and play by its own rules.

    Other empires are pretty central, unfortunately. Just the desolation monsters won’t be enough tension really and will be too purely defensive… although maybe. I’ll think about it!

  • Widminter

    There is one, and only one, difficulty option I’d like to have in almost any game with an element of reaction time – be they a shooter, real-time strategy, racer, platformer or whatever. The option I’d like is to change the overall game speed. As you say, don’t change the number of hostiles, or their hit points, or the player stats. Just give me the option to acknowledge that I simply don’t have the reactions of a 14-year-old any more. Meanwhile, the actual 14-year-olds can post speedruns playing in caffeine mode.

    This is frequently implemented in real-time strategy games. I’ve seen it in a few others, like One Must Fall: 2097 (2D fighter) and Tyrian (side-scrolling shooter). Some emulators also allow you to adjust the speed, which is nice.

    I’m certainly frustrated when “difficulty levels” don’t address the issues that make a game difficult. I’ve seen many platform the “easy” mode reduces the number of enemies or increases my hit points, but the thing that actually kills me is pits or other instant-kill environmental hazards. In reverse, sometimes I find a game tedious, so I try a higher difficulty – and it just fills the levels with more or tougher enemies, making the game more tedious.

  • keithburgun

    I’m OK with this but only if it’s made VERY CLEAR what the optimal intended gameplay speed is.

  • Jeremy Watson

    I like your discussions very much. Difficulty itself is defined as imbalance? If you don’t change numbers or just cheat outright, just by making the AI select bad choices on purpose you are losing something.

    I will admit though having a design that improves AI choices in the game mechanics without cheating is far superior design to me than just adding stats or abilities. A more difficult opponent in PvP games would make smarter choices, not just handicap to gain an advantage.

    Starcraft AI actually does both, on elite difficulty the AI gains a huge artificial economy boost, but will also pull workers off the line to prevent rushes, scout and counter-build. In this case, I just don’t know how they could have done without the economy boost. What would you suggest?

  • keithburgun

    I’m not sure why you’re talking about AI so much; it doesn’t seem like it’s the same topic as my article? But anyway, I think that AI is basically just a bad idea overall in terms of game design. If your game is reliant on AI, then really, it’s a puzzle. We just have to figure out how to exploit the AI, and then it’s SOLVED. So I oppose the idea of using AI at all (beyond INCREDIBLY simple AI that isn’t supposed to be smart, like in AURO, which I would barely even consider AI).

  • Jeremy Watson

    You’re absolutely right I did miss… I guess I just jumped to AI because that is another tool to augment difficulty and I guess I got my answer as to where you stand on it. Personally I don’t care about difficulty either.

    I find AI done somewhat well very interesting, but off topic.

  • GamingWolf

    I appreciate your radical insight rarely seen and quite academic analysis of your personal topics in regards to game design but I believe you have poisoned the well when it comes to semantics so much that are you attempting to control the narrative entirely. The term, ‘difficulty’ is just another word on your list of concepts you consider myths. This wouldn’t even be an issue if you weren’t attempting to rig public opinion and appealing to higher and higher institutions.

    I miss when you were an underdog of gaming opinion. Invent new words if you have to as these concepts are quite broadly defined.

    Also going by even older games now I’m not even sure what type of game would appease you now. Why not create something completely different altogether?

  • keithburgun

    I think older games were just a different kind of bad, really. The best videogames I’ve played are OUTWITTERS on iOS, which completely pleases me, and Desktop Dungeons. Super Smash Brothers 64 is very good. Many more boardgames like Puzzle Strike, Through the Desert, Puerto Rico, etc.

    >Why not create something completely different altogether?

    That’s what I am doing with AURO and Empire.

  • GamingWolf

    Well thanks for replying, I needed to get that off my chest.

    (I love Desktop Dungeons btw)

  • Sq

    I think difficulty levels are possible, they just take nearly as much work as designing an entirely new game, rather than the “slap a hitpoint scale on it” approach used more typically. The easiest (which is still harder than what’s usually done) way to implement difficulty that I know of is to increase the number of abilities the player has as difficulty goes up, and scale the challenges to optimal use of all abilities. That is, you always have to do close to the best you can with the tools the game provides you, but keeping up with that standard becomes harder as the game gives you more tools. Or, in a concrete example, imagine a Starcraft map where you have to destroy as many enemy forces as possible in a time limit. On easy mode, you’re only allowed to build zerglings, drones, and hatcheries, and you face challenges that can be overcome with these. On normal mode, you’re also allowed infestors, but face challenges that make them necessary. And so forth.